Fireside corrosion degradation of 15Mo3, T22, T23 & T91 in simulated coal-biomass co-fired environment

Transcription

1 Materials and Corrosion 2015, 66, No. 9 DOI: /maco Fireside corrosion degradation of 15Mo3, T22, T23 & T91 in simulated coal-biomass co-fired environment T. Dudziak*, T. Hussain, D. Orlicka, A. Pokrywa and N.J. Simms This paper reports the result of a fireside corrosion test carried out on four low alloyed ferritic alloys at 650 8C for 1000 h. The combustion gases were derived from co-firing cereal co-product with a UK coal. Synthetic deposits were used to assess the influence of the deposit compositions. The samples were examined using an ESEM/EDX. To quantify the metal damage the pre-exposure micrometre measurements were compared with the post-exposure image analyser measurements on sample cross-sections. The median metal damage of the alloys showed the following ranking (most to the least damage): 15Mo3>T22> T23> T91. 1 Introduction One of the major issues facing the world is climate change, which is believed to be due to the increase in greenhouse gas emissions. According to the International Energy Agency (IEA) energy statistics, electricity and heat production accounts for 41% of the total CO 2 emission [1 3]. Conventional solid fossil fuel power plants contribute significantly to this CO 2 emission and the EU has put in place legislation to reduce its emission to 20% below the 1990 levels by The UK also has an ambitious target of reducing CO 2 emissions to 80% of their 1990 levels by 2050 [4]. To meet this ambitious target a combination of renewable fuels such as biomass and highly efficient ultra supercritical power plants with carbon capture and storage technology (CCS) will be necessary. Biomass co-firing with coal at a lower percentage in conventional pulverised fuel power plants has proved to be a successful way to reduce the overall CO 2 emission from electricity generation industry. Biomass accounts for approximately 70% of total renewable energy production in the world [5]. Biomass absorbs CO 2 during its lifetime and releases the CO 2 during the combustion; hence, it can be classed as T. Dudziak Foundry Research Institute in Cracow, Zakopiańska , Kraków (Poland) tomasz.dudziak@iod.krakow.pl T. Hussain Division of Materials, Mechanics and Structure, University of Nottingham, Nottingham, NG7 2RD (UK) D. Orlicka, A. Pokrywa, N.J. Simms Institute of Energy and Resource Technology, Cranfield University, Cranfield, MK43 0AL (UK) This study has been performed in Cranfield University at Institute of Energy and Resource Technology, MK43 0AL, (UK). carbon neutral. Dedicated biomass fired plants are usually smaller in size (and lower overall efficiency) and the plant can face significant issues with fouling, deposition, agglomeration and corrosion. These issues restrict the overall steam temperature that can be used in those plants. However, burning a smaller percentage of biomass with coal in existing pulverised fuel power plants, which are much larger in capacity and more efficient than dedicated biomass fired plants, can produce more biomass derived renewable energy. Reduction in CO 2 emissions from pulversied coal fired power plants can be achieved by increasing the operating temperature (and pressure) of the steam systems, which results in an increase in overall plant efficiency. Typically 1% increase in absolute efficiency results in as much as 3% reduction in CO 2 emissions [6]. Fireside corrosion has been an on-going concern for the power generation industry where coal has been used as a fuel. Fireside corrosion is defined as the loss of the heat exchanger material due to chemical reactions with the surrounding environment at high temperature. This can lead to tube failures, which are expensive to repair and unplanned shutdown which can cause large financial losses. In addition, future ultra-supercritical/supercritical plants will operate at steam temperatures beyond 700 8C to increase the overall efficiency in power generation [3]. Increased operating temperatures coupled with biomass derived fuel gases will exacerbate the fireside corrosion problem. Typically, the fireside corrosion shows characteristic bell-shaped curves, where the corrosion rate increases with increasing temperature and then decreases as the deposit becomes unstable. It has been reported that the peak corrosion damage in air-fired combustion gases occurs around 650 8C [7 9]. Therefore, it is important to investigate the fireside corrosion performance of the ferritic alloys at that temperature. Assuming the fireside corrosion mechanism operating in the temperature ranges C remain the same, increasing the temperature would also accelerate the corrosion rates which will provide valuable information about wileyonlinelibrary.com 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

2 840 Dudziak, Hussain, Orlicka, Pokrywa and Simms Materials and Corrosion 2015, 66, No. 9 the lifetime of the heat exchanger materials. Furthermore, it is important to investigate the influence of various chemical compositions in the deposit on the fireside corrosion mechanisms. Therefore this paper reports the results of the influence of different chemical compositions of the deposits on the ferritic steels in simulated fireside corrosion tests. The laboratory tests were conducted with 15Mo3, T22, T23 and T91 materials in simulated air-fired combustion gases at metal temperature higher than average conventional operating temperatures of the existing pulverised fuel power plants. The gas and deposits were selected based on co-firing a UK coal with a cereal co-product (CCP). The performance of the alloys was investigated with five different deposits. The work was carried out according the deposit-recoat test method that has been developed for high temperature corrosion tests and was used in recent years [8,10]. Dimensional metrology has been used as the primary route to quantify the metal damage occurring due to fireside corrosion. The exposed samples were examined in environmental scanning electron microscope (ESEM) in backscattered electron mode (BSE) equipped with energy dispersive X-ray spectrometry (EDX) to characterise corrosion degradation. 2 Experimental procedure 2.1 Materials In this study four alloys were used: a low carbon ferritic steel with no Cr (15Mo3), two ferritic steels with 2.3 wt.% Cr (T22 and T23), and one ferritic/martensitic steel with 9 wt.% of Cr (T91). The materials were chosen to investigate the mechanism under different deposits in air-fired conditions. The materials represent materials currently used for the superheater/reheater tubes (SH/RH) in pulverised fuel power plants. The nominal chemical compositions of the materials are shown in Table 1. Each of these materials was machined from commercial boiler tubes of various dimensions. The shape of the samples used in this study is shown in Figure 1. 15Mo3 was machined as rectangular section (due to lack of tube segments) and the rest of the materials were machined as tube segments. The typical dimensions of specimens were 15 mm long, 10 mm wide/ chord (for tube sections) with a 4 mm wall thickness. All surfaces of these tubes were machined to a UK 600 grit surface finish (Ra <0.4 µm). Figure 1. Geometry of the exposed materials: A) 15Mo3, B) T22, T23 and T Gases and deposits The exposure conditions were determined following a detailed investigation of the gaseous environments and deposit conditions that could be found around heat exchangers (superheaters/reheaters) in conventional pulverised fuel power plants using a UK daw mill CCP [7,10,11] as mentioned previously in this study. The co-firing ratio of the biomass was 20 wt.% for this study, the compositions of the fuels are available in ref. [7]. The gas composition from the combustion of these fuels were calculated using pre-existing models and simplified to their active components for corrosion testing in the controlled atmosphere furnaces. The nominal gas compositions used in the corrosion tests are shown in Table 2. The alloys were exposed with five synthetic deposits in this test programme. Table 3 shows the composition of the deposits. The deposits in the experiment were chosen to simulate the deposits formed on the heat-exchanger surfaces in the pulverised fuel power plants. In this study in total 24 samples were used, 4 samples without deposit (D0) and another 20 specimens with 5 different deposits (D1 D5). The deposits can be characterised as follows: D0: no deposit (i.e., bare alloy surface exposed to the gas atmosphere). D1: a standard deposit composition that is widely used in screening tests; it represents a composition of alkali iron trisulphate that has been identified from many investigations as the principal cause of fireside corrosion in superheaters/reheaters in coal-fired power stations [9,11,12]. D2: the alkali iron tri-sulphate compositions from D1 diluted with kaolinite (Al 2 O 3.2SiO 2.2H 2 O) to represent the clay minerals usually found in coals along with Fe 2 O 3 and CaO. K 2 SO 4 concentration was decreased to 0%, whereas Na 2 SO 4 was reduced to 10% in order to minimise aggressiveness of the deposit D2. D3: is similar to D2 but with equal amount of K 2 SO 4 and Na 2 SO 4 (5%) to increase the aggressiveness of the deposit, other Table 1. Nominal compositions (wt.%) of the materials used in the fireside corrosion test. Alloy Fe C Mn P S Si Cr Ni Mo V W Nb B Al N Cu 15Mo3 Bal T22 Bal T23 Bal T91 Bal WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

3 Materials and Corrosion 2015, 66, No. 9 Fireside corrosion degradation of 15Mo3, T22, T23 & T Table 2. Nominal composition of the mixture of gases when used in the corrosion test O 2 % CO 2 % H 2 O% N 2 % SO 2 vpm HCl vpm Bal elements (kaolinite, Fe 2 O 3 and CaO) are in the same concentration as shown in D2. D4: is similar to D3, but with more KCl (10%) to investigate the sensitivity of the corrosion damage as a result of biomass co-firing. D4 does not contain any K 2 SO 4 ;however Na 2 SO 4,Fe 2 O 3 and CaO phases are in the same amounts as in D3. D5: is similar to D2, with the same amount of Na 2 SO 4 (10%). Hematite (Fe 2 O 3 ) concentration in both deposits is the same; however, D5 does not contain any CaO (burnt lime). 2.3 Experiment setup The fireside corrosion exposures were carried out in an alumina lined vertical controlled-atmosphere furnace using simulated airfired combustion gases. A schematic diagram of the fireside corrosion tests is shown in Figure 2. The furnace accommodates 24 test pieces in alumina crucibles at one time in the hot zone. Each test was run for 1000 h using the widely accepted deposit-recoat method [8]. Prior to sample exposure, the samples were cleaned using a degreaser (volasil) followed by isopropanol in an ultrasonic bath at a temperature of 35 8C for 20 minutes. For those samples being exposed with a surface deposit, the cleaned samples were painted using a paintbrush to apply a deposit loading of 20 mg/cm 2. The test was cycled every 200 h and fresh deposits were applied on the exposed samples in order to replenish the salts resulting in a deposition flux of 100 mg/cm 2 / h (no deposit was removed before painting in this widely accepted deposit-recoat test method). As part of the depositrecoat process (5 cycles), the samples were weighed every 200 h with and without crucibles as well as before and after applying the deposits. Premixed gases were supplied to the controlled-atmosphere furnace through mass flow controllers to achieve the desired gas composition. The gas containing (CO 2, O 2, N 2 ) was passed Figure 2. A schematic diagram of the fireside corrosion test setup through a de-ionized water bubbler, which was kept at 40 8Cina water bath to add the required amount of moisture to the gas stream (8.6 vol.%) before mixing with the corrosive species (HCl, SO 2 ). 2.4 Sample metrology: pre- and post-exposure measurements Following the exposures, the samples were vacuum mounted using a low shrinkage cold mounting resin filled with ballotini (to further reduce shrinkage) in a bespoke jig. The mounted sample was cross-sectioned, ground and polished to 1 µm using non-aqueous lubricant. ESEM fitted with an EDX system was used to investigate the scale/deposit thickness and microstructure. The dimensional metrology of the samples is a key feature of this test programme. The dimensions of each sample were measured before the exposure using a micrometre (with resolution of 1 µm). Post-exposure measurements were carried out using a LEITZ WEIZLAR optical microscope coupled to an IMAS image analysis system with a calibrated x-y stage capable of positioning the sample to an accuracy of 2 mm. The system can automatically take groups of pictures from locations evenlyspacedaroundthesample outer edge. From these pictures, positions of the outer sample surface, and depth of internal damage are determined (using x-y coordinates of the features of interest relative to a reference point for whole sample). By comparing sample dimensions before and after exposure, the apparent metal Table 3. Deposits used in the simulated air-fired experiments (mol%) Deposit Kaolonite Fe 2 O 3 CaO CaCO 3 Na 2 SO 4 K 2 SO 4 KCl D D D D D D WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

4 842 Dudziak, Hussain, Orlicka, Pokrywa and Simms Materials and Corrosion 2015, 66, No. 9 loss and the change in sound metal (metal loss þ internal damage) can be calculated. These data sets can then be re-ordered (from greatest to least metal loss) and corrected for calibration differences (using data from reference samples). The processed data can then be plotted as a metal loss vs. cumulative-probability. This measurement method was performed in accordance with the draft standard methods for high temperature corrosion assessment [11]. 3 Results 3.1 Kinetics Mass change data is the most conventional and frequently reported method of observing metal oxidation and corrosion at high temperatures. Figures 3 8 illustrate the specific net mass change [mg/cm 2 ] vs. time [h] of the exposed samples: 15Mo3, T22, T23 and T91 with deposits D0 D5 after 1000 h of exposure at C. 15Mo3 shown in Figure 3 showed the highest mass gain when bare alloy was exposed and the lowest rate of corrosion was observed in T91 where concentration of Cr was the highest. The alloys T22 and T23 (with similar Cr content) showed similar kinetic behaviour. It is difficult to rationalise the mass change data of the alloys with deposits in high temperature fireside corrosion. Nonetheless, the mass change data for the alloys under all five deposits are presented in this paper for comparative purposes. Figure 4 shows mass gain of all alloys used in this study under D1 deposit. The results show much higher mass gain under deposit D1 than under other deposits used in this work. Ferritic alloys under D1 deposit could not form a protective oxide scale to improve its corrosion behaviour. Kinetic behaviour showed linear characteristics, where mainly chemical reactions at the deposit/oxide scale interface and the oxide scale/substrate interface determined the performance. The highest mass gain was seen by T22 alloy, the lowest mass gain was seen by T91 alloy. Such behaviour indicates the amount of Cr in the alloy did not correlate to their performance under the aggressive deposit. Figure 4. Mass gain of the alloys exposed with D1 deposit at Cin air-fired conditions There was no clear distinction in the mass gain data of the alloys under deposits D2, D3, D4 and D5 shown in Figures 5 8 respectively. The alloys covered by the deposits D2 D5 were rich in kaolinite (alumina silicate found in coal ash) and much lower corrosion degradation was observed compared to deposit D1. Typically, the lowest corrosion rate was found in T91 alloy which contains 9 wt.% of Cr and the highest in 15Mo3. Figure 5. Mass gain of the alloys exposed with D2 deposit at Cin air-fired conditions Figure 3. Mass gain of the alloys exposed with no deposit (D0) at C in air-fired conditions Figure 6. Mass gain of the alloys exposed with D3 deposit at Cin air-fired conditions 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

5 Materials and Corrosion 2015, 66, No. 9 Fireside corrosion degradation of 15Mo3, T22, T23 & T Figure 7. Mass gain of the alloys exposed with D4 deposit at Cin air-fired conditions 3.2 Microstructural investigations The surface observations were conducted only for the alloys with no deposits; however, due to the similarity in microstructures micrographs are not shown here. Alloy 15Mo3 showed needle like structure growing uniformly across the surface, which suggests a highly active surface going through rapid oxidation/ corrosion degradation. The oxide growing on the T22 alloy was not uniform and had several bald patches. In contrast, the surface of the T23 did not show the needle like structure. The alloy with the highest Cr content (T91) showed irregular oxide growth similar to the T22; however the shapes of the oxides were well defined. It can be concluded that all four materials underwent corrosion degradation in the simulated co-fired combustion gases without any deposit. In general, porous and non-protective scales developed on the surfaces of the exposed alloys. The analyses performed by EDX confirmed that mainly Fe 2 O 3,Fe 3 O 4 and some amounts of sulphur was found in the interface oxide scale/substrate. The higher amount of Cr did not result in a non-porous and uniform oxide growth on the surface. The samples were crosssectioned to identify the composition of the mixed oxidescorrosion products. The cross-sectioned microstructural analysis was performed on the materials with deposit D1 (most aggressive) and without Figure 8. Mass gain of the alloys exposed with D5 deposit at Cin air-fired conditions deposit D0 (no deposit) only in order to study the scale/corrosion product morphology and degradation after exposure. The alloys with deposit D2 D5 were as well investigated using SEM/EDX and optical microscope; however, these observations are not shown here. The purpose of this work was to focus on the most corroded samples and the samples with no deposit in order to compare and evaluate corrosion rate to establish a rank order of alloys. Figures 9A 9D show the cross sectioned images of all four exposed materials after 1000 h in simulated co-fired combustion gases with no deposit (D0). The BSE images indicate that all four materials underwent a high degree of corrosion as expected. Thick (500 µm) layer of oxide scale was formed on 15Mo3, which delaminated from the substrate (either during sample preparation or thermal cycling). The oxide scale consisted of Fe 2 O 3 (top layer 150 µm) and inner part was occupied exclusively by Fe 3 O 4 with additional amounts of sulphur (up to 1.8 wt.%), thickness of the scale reached 350 µm. The scale consisted of trace amount of Mn at the oxide scale-substrate interface.the oxide scale on T23 did not show any lateral cracks like the scale grown on the T22. However, the scale on the T22 showed higher surface roughness (Ra) at the scale-metal interface. The scale grown on the T23 is a multi-layered structure slightly enriched in the interface by Cr and W diffused from the bulk alloy. The lack of adherence could be enhanced by thermal stresses within the oxide scale upon heating and cooling. The alloy with the highest Cr content (T91) showed very thin oxide scale formation (100 µm); however, this oxide scale detached from the substrate either due to the high stress concentration level upon cooling or during sample preparation. Figures 10 and 11 show phase stability diagrams for Fe S O and Cr-S-O respectively in air fired conditions. The stability phase diagrams were calculated using FACTSAGE software [13]. The stability phase diagrams show that Fe 2 O 3 and Fe 3 O 4 are stable under air-fired conditions. The test condition used in this study is also identified in the figure. At the higher partial pressure of SO 2, FeS and FeS 2 phases are thermodynamically stable. Cr-S-O phase diagram indicates formation of Cr 2 S 3 at high partial pressure of SO 2 and a low partial pressure of O 2, which could form only at the interface of oxide scale and the substrate. This phase could form only in the alloy with the highest Cr content (T91) where activity of Cr enhances the formation of Cr 2 S 3.Cr 2 O 3 is stable over a large range of partial pressure of O 2. Figures 12A 12D show the cross section images of 15Mo3, T22, T23 and T91 after 1000 h exposure with a screening deposit D1 (Na 2 SO mol.%, K 2 SO mol.%, Fe 2 O 3 25 mol.%). All the four alloys showed similar scale morphologies under D1 deposit. It was found that Cr level did not play an important role in the corrosion protection of the alloys when covered with D1 deposit. The level of Cr in the alloys could not form a protective oxide layer under the test conditions with such an aggressive deposit. The exposed samples underwent high degradation rate, where thick nonprotective scales formed. Na, S and K from D1 deposit diffused towards the oxide scale interface and Cr diffused outwards from the exposed alloys to form the oxide scales WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

6 844 Dudziak, Hussain, Orlicka, Pokrywa and Simms Materials and Corrosion 2015, 66, No. 9 Figure 9. Cross sectioned images in the BSE mode of: A) 15Mo3, B) T22, C) T23 and D) T91 alloys exposed in air-fired condition at C for 1000 hours without deposit (D0) Morphologically, the oxide scales are very similar despite different Cr content in the base alloys - the alloys mainly developed Fe 3 O 4.Fe 2 O 3 was not found within the oxide scale but a small amount of Fe 2 O 3 phase was found in mixed deposit scale. 3.3 Metal loss data of the exposed materials Dimensional metrology provides the most reliable measurement of the corrosion damage to the alloys in fireside corrosion. It produces a distribution of metal damage data for each exposed sample. The results are plotted as metal loss vs. cumulative probability according to the draft standard for high temperature corrosion measurements [8,11]. Figures 13A 13D show the metal loss data of all the four exposed materials at C after 1000 h of exposure in air-fired conditions with all deposits (D0-D5). The alloys cover with deposit D1 showed the highest corrosion degradation in all four alloys. A useful way of comparing the data is to look at the median (50% cumulative probability) metal loss. For example, Figure 13 shows that the median metal loss of 15Mo3 was 580 µm with deposit D1 and Figure 10. Phase stability diagram for Fe-S-O in air fired conditions at C; black dot represents an indication of experiment conditions in partial pressure of SO 2 and O 2 Figure 11. Phase stability diagram for Cr S O in air fired conditions at C; black dot represents an indication of experiment conditions in partial pressure of SO2 and O WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

7 Materials and Corrosion 2015, 66, No. 9 Fireside corrosion degradation of 15Mo3, T22, T23 & T Figure 12. Cross sectioned images in the BSE mode of: A) 15Mo3, B) T22, C) T23 and D)T91 alloys exposed in air-fired condition at C for 1000 hours with deposit 1 (D1) Figure 13. Change in metal thickness versus cumulative probability showing the fireside corrosion behaviour of: A) 15Mo3, B) T22, C) T23 and D) T91 alloys in this test with D0 D5 deposits for 1000 hours at C WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

8 846 Dudziak, Hussain, Orlicka, Pokrywa and Simms Materials and Corrosion 2015, 66, No µm with the rest of the deposits. The rates are excessive compared to the traditional target corrosion rate value for superheater/reheater materials in conventional power plants, which is around µm/1000 h. It should be noted that deposit D1 is an aggressive deposit mainly used for screening tests. Corrosion degradation of all the four alloys using deposit D0, D2- D5 were mostly uniform as can be seen from the lack of a step change in the cumulative probability curves. However, this was not the case with the T23 and T91 alloys exposed with deposit D1. Figures 13 C- D show that 18% of the T23 sample surface suffered from a corrosion damage in excess of 400 µm and similarly 16% of the T91 sample surface suffered from a corrosion damage in excess of 280 mm with deposit D1. This is due to the formation of localised pits on the sample surface. A strong relationship between the corrosion degradation with the Cr content of the alloy was observed in the dimensional metal loss data. The lowest degradation was seen in T91 and the highest degradation was seen in 15Mo3 (looking at the median metal damage). The median metal loss vs. Cr content of the alloys (increasing order) with all the deposits are shown in Figure 14. The metal loss values have been multiplied by 1 and presented as metal loss in this figure. The figure shows that the median metal loss values decreased with increasing the Cr content of the alloy; for example with deposit D1 the following median metal loss values were observed: 570 mm in 15Mo3, 400 mm in T22, 320 mm in T23 and 280 mm in T91. It is worth pointing out that T23 outperformed T22 in these test conditions. Both T22 and T23 contain similar amount of Cr but T22 contains 1% Mo and T23 contains 1.5% W and no Mo. 4 Discussion Discussion in this work is divided into subsections; subsection 4.1 is focussed on the alloys exposed to fireside corrosion regime with no deposit, subsection 4.2 describes the corrosion mechanisms induced by deposit D1, subsection 4.3 discusses the role and mechanism under deposit D2, subsection 4.4 focuses on deposit D3, subsection 4.5 discusses the role of deposit D4, finally subsection 4.6 shows outlook on deposit D Alloys with no deposit The study showed that ferritic steels 15Mo3, T22 and T23 exposed to fireside conditions without deposit generally developed thick and non-protective scales. Lack of adherence between the layers in the formed oxide scale in ferritic steels with low amount of Cr can be attributed to the unbalanced fluxes across the scale and unbalanced diffusional processes known as Kirkendall effect [14]. The morphology in 15Mo3 alloy developed the scale consisting Fe 2 O 3 (top layer) and Fe 3 O 4 (inner layer) only with additional amount of sulphur (1.7 wt.%). Ferritic steel with 2.3 wt.% of Cr showed better morphology than that observed in 15Mo3, the oxide scale more adherent where no spallation or detachment was observed. The scale consisted of Fe 2 O 3 at the top and Fe 3 O 4 in the middle part of the oxide scale with small amounts of sulphur 0.5 wt.% and chromium 2.5 wt.% at the interface. The oxide scale formed on T23 alloy showed similar phase composition where top layer was composed of Fe 2 O 3 and inner layer was composed of Fe 3 O 4 with up to 2.5 wt.% of Figure 14. Median change in metal loss of 15Mo3, T22, T23 and T91 alloys after 1000 h at 650 8C in simulated co-fired combustion gases with deposits (D0-D5) 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

9 Materials and Corrosion 2015, 66, No. 9 Fireside corrosion degradation of 15Mo3, T22, T23 & T sulphur, chromium (2.8 wt.%) and tungsten (2.6 wt.%). It is believed that W is added to T23 in order to support Cr diffusion to the oxide scale during the high temperature exposure. The mechanism can be explained through the chemical reaction where C from the alloy forms WC carbide and further formation of carbide decreases the activity of both C and W thus Cr activity can be increased. and releases SO 3, which diffuses to the cooler part of the material, while a layer of the slag forms on the outer surface. The chemical reaction of the formation of alkali iron tri-sulphates occurs through the following way [9]: SO 2ðgÞ þ 1 2 O 2 ðgþ! SO 3ðgÞ ð2þ W þ C! WC ð1þ Fe 2 O 3ðsÞ þ 3SO 3ðgÞ! Fe 2 ðso 4 Þ 3 ðsþ ð3þ The addition of W to T23 alloy should enhance the activity of Cr to form more protective scale than that formed in T22 alloy where Cr from the bulk alloy is directly consumed for the reaction with C in order to form carbides such as Cr 3 C 2,Cr 7 C 3, and Cr 23 C 6 [15], which decreases the Cr activity resulting in formation of thicker non protective scales. Nevertheless, both low-alloyed steels T22 and T23 formed thick scales in this study. T91 alloy with the highest Cr content, showed the formation of thin non adherent oxide scale with following phases: top layer was occupied by Fe 2 O 3, inner layer was occupied by Fe 3 O 4, the inner most layer was occupied by Cr 2 S 3. The layer of Cr 2 S 3 according to Huczkowski et al. [16] formed due to the inward diffusion of SO 2 from ambient atmosphere through the porous oxide scale. The formation of Cr 2 S 3 is due to the very low partial pressure of oxygen and high partial pressure of sulphur. Thus in such conditions, it is more likely to form Cr 2 S 3 than Cr 2 O 3. The high activity of Fe driven by high concentration and inward diffusion of sulphur from ambient atmosphere can form Fe-S phase; however, Cr 2 S 3 is more stable and the level of sulphur within the oxide scale decreases constantly, thus FeS formation is not favourable. In contrast to Huczkowski et al. [16], Stringer et al. [17] concluded that patches of FeS can be observed in the deep regions in the interface of the oxide scale and ash deposit. 4.2 Alloys with D1 deposit All four alloys formed thick 1 mm scales, suggesting significant corrosion damage under aggressive deposit D1 (Figure 12). It is believed that observed degradation was due to the formation of alkali iron tri-sulphates, which was detected in XRD scans on the surface of the samples following exposure [18]. Chemical composition of the deposit D1 used in this work was designed to invoke the formation of alkali iron tri-sulphates [8,12] and the EDX analyses of the exposed samples following exposure also showed compositions similar to alkali iron tri-sulphates [10,19] [20]. Moreover, alkali iron tri-sulphates were reported in other studies where deposit composition similar to the composition of D1 was used [20]. According to Natesan et al. [9] and later Karlsson et al. [21] corrosion induced by alkali iron tri-sulphates begins from the formation of thin layer on the metal surface, or by the oxidation of Fe-S phase in coal to form iron oxide in SO 2 /SO 3 gases. Alkali sulphates with oxides of sulphur in the boiler are deposited over the oxide scale on the exposed materials.the outer surface of the deposit due to temperature gradient becomes sticky and particles of ash are captured. Increasing temperature of the deposit induces thermal dissociation of sulphur compounds in the ash 3ðNa=KÞ 2 SO 4ðlÞ þ Fe 2 ðso 4 Þ 3ðsÞ! 2ðNa=KÞ 3 FeðSO 4 Þ 3ðs;lÞ ð4þ Corrosion degradation increases with increasing temperature in the presence of molten deposits, below C corrosion rate in low ferritic steels is very low (1.5 mm per year), at C corrosion rate increases two orders of magnitude (350 mm per year) and 3000 mm per year corrosion rate is observed when temperature reaches melting point of alkali iron tri-sulphates at around C [22]. The temperature depends on the chemical compositions and SO 2 /SO 3 balance in the atmosphere [23,24]. Corrosion rate decreases as the deposit becomes unstable due to a shift in the balance of SO 2 /SO 3 towards SO 2 at higher temperatures. The formation of unstable (in this conditions) Fe 2 (SO 4 ) 3 reacts further with Na and K to form alkali iron tri-sulphates. 4.3 Alloys with D2 Deposit D2 contained much lower concentration of sodium and potassium sulphate than that observed in deposit D1 and similar amounts of hematite. Reduced amounts of sulphates were replaced by kaolinite in order to reduce aggressivess of the deposit. The results achieved in this study clearly show that lower corrosion degradation was observed. It is believed, that lower corrosion rate was reduced by addition of kaolinite to the deposit D2. Addition of kaolinite reduced concentration of sodium and potassium sulphates, as a result much lower corrosion degradation was observed. The addition of CaO (burnt lime) combined with Al 2 O 3.2SiO 2.2H 2 O (kaolinite) resulted in reduction of active alkali flux in the deposit. At high temperatures kaolinite underwent series of transformations. Endothermic dehydration reaction due to the large amount of energy required to remove the chemically bonded hydroxy ions starts at temperature range C. As a result, kaolinite dehydration of kaolinite produces disordered meta kaolinite Al 2 Si 2 O 7 according to the following reaction [25]: 2Al 2 SiO 5 ðohþ 4 ðsþ! 2Al 2 Si 2 O 7 ðsþ þ 2H 2 O ðgþ ð5þ The addition of CaO (burnt lime) in D2 does not play an important role due to the low concentration; however, 5% of CaO represents characteristic value for the fly ash derived from coal burning process. CaO plays an important role in slagging and deposit formation, where very high fusion temperatures tend to inhibit the eutectic effect of alkalis. Reducing sodium WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

10 848 Dudziak, Hussain, Orlicka, Pokrywa and Simms Materials and Corrosion 2015, 66, No. 9 sulphate concentration in D2 compared to deposit D1 decreased the rate of corrosion degradation where much lower mass gain was observed. There were no potassium species present in the deposit D2 and hence only sodium iron trisulphate was formed exclusively according to the reactions (2) (4). The formation of alkali iron tri-sulphate was described in reaction (4), alkali iron tri-sulphate formed beneath the layers of sticky ash where oxygen partial pressure was much lower than sulphur. 4.4 Alloys with deposit D3 The alloys covered with deposit D3 underwent similar corrosion rate as those covered by D2, however chemistry of D3 was slightly different than observed in D2. Deposit D3 contained 5% of K 2 SO 4 where in D2 deposit K 2 SO 4 was not present. Presence of K 2 SO 4 should increase corrosion degradation of low alloyed steels, however as the study showed kinetics and metal loss were similar to those obtained under D2 deposit. The exposure of the low alloyed steels invoke several reactions at high temperature. Firstly, at high temperature the alloys formed the scale consisted of Fe 3 O 4 (inner layer) and Fe 2 O 3 (top layer), further, deposit D3 started to react with the formed oxide scale with gaseous elements such as HCl to form MCl, where M (K, Na or Fe). Aho et al. [26] proposed the following mechanism: 2MCl þ SO 2ðgÞ þ 1 2 O 2 ðgþ þ H 2 O ðgþ! M 2 SO 4 þ HCl Al 2 O 3 SiO 2 þ 2MCl þ H 2 O! M 2 O þ Al 2 O 3 SiO 2 þ 2HCl ð6þ ð7þ Further, Aho et al. [26] concluded that depending on the initial concentration of alkali, sulphur, kaolinite, chlorine cocombustion favoured the inhibition of chlorine deposition especially by the mechanism shown above. This mechanism can be confirmed in this study as well, In spite of the addition of 5 wt.% K 2 SO 4 in D3 did not change the overall degradation mode. Moreover, it can be assumed that the high concentration of kaolinite and CaO acted as a buffer. The following reactions can take place during high temperature exposure in air-fired conditions. The formed oxide scale consisting mainly Fe 3 O 4 reacts with K 2 SO 4 and Na 2 SO 4 similar to reaction (4), whereas kaolinite decompose to meta kaolinite according to reaction (5) shown previously. At the same time, CaO can react with K 2 SO 4 and form calcium sulphate and potassium oxide respectively: Kinetic data and metal loss data obtained in this work clearly shows that both sodium and potassium calcium tri-sulphate Na 2 Ca 2 (SO 4 ) 3 and K 2 Ca 2 (SO 4 ) 3 respectively are far less corrosive than K 3 Fe(SO 4 ) 3 and Na 3 Fe(SO 4 ) 3 formed under deposit D1. The bio fuels with high annual growth, such as fast-growing woods, forest residues, annual crops and straws, all have abundant alkali in the ash where potassium, in particular volatilizes and reacts readily during combustion. Wood fuels contain approximately 0.1% elemental potassium (K) compared with 1% or higher in straws and grasses, hulls, pits and shells [27]. 4.5 Alloys with deposit D4 Deposit D4 was introduced in this study in order to investigate the sensitivity of corrosion damage as a result of biomass cofiring. Thus the mechanism of corrosion under D4 deposit can be explained in the following way: the alloy with low concentration of Cr in the matrix is unable to form a protective oxide scale and thus undergoes corrosion degradation according to chlorine mechanism, where firstly HCl (400 ppm) decomposes at high temperature to H 2 and Cl 2 : HCl! H 2 ðgþ þ Cl 2 ðgþ ð11þ Due to the decomposition of HCl concentration of both hydrogen and chlorine increase. Hydrogen can cause hydrogen embrittlement of the oxide scale [20,21] whereas chlorine further accelerates degradation rate of the low-alloyed steels by means of chlorine effect. According to Nieken et al. [28] mechanism can be rationalized by the Deacon reaction: HCl þ 1 2 O 2! Cl 2 ðgþ þ H 2 O ðgþ ð12þ The reaction can be further accelerated by the oxidised surface. According to the thermodynamics of the Metal-Cl-O system, metal oxides are stable at high oxygen partial pressures while metal chlorides can be expected in regions with low oxygen partial pressure. Thus, such conditions are expected at the surface of the oxide scale where oxygen partial pressure is higher than that observed underneath deposit. The metal chlorides are usually observed at the metal-scale interface (low po 2 ), while the outer part of the scale (high po 2 ) contains mainly metal oxides [29,30]. It is also known that chlorine diffuses inwards through the oxide scale of Fe 3 O 4. At the metal/oxide interface the oxygen pressure is low and metal chlorides are stabilised according to the following reaction: K 2 SO 4 ðsþ þ CaO ðsþ! CaðSO 4 Þ ðsþ þ K 2 O ð8þ Me x þ 1 2 Cl 2! MeCl x ðs;gþ ð13þ Further CaSO 4 reacts with both sulphates Na 2 SO 4 and K 2 SO 4 according to the following reaction: 2CaSO 4ðsÞ þ K 2 SO 4ðsÞ! K 2 Ca 2 ðso 4 Þ 3ðsÞ ð9þ The vapour pressure of the chlorides at elevated temperature reach significant values and volatile chlorides diffuse towards oxygen rich outer part (oxide scale) where it undergoes destabilisation [29,30]. 2CaSO 4ðsÞ þ Na 2 SO 4ðsÞ! Na 2 Ca 2 ðso 4 Þ 3ðsÞ ð10þ 2MeCl 2 þ 3 2 O 2! Me 2 O 3ðsÞ þ 2Cl 2 ðgþ ð14þ 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

11 Materials and Corrosion 2015, 66, No. 9 Fireside corrosion degradation of 15Mo3, T22, T23 & T Due to the formation, decomposition reaction in this model, oxide scale formed on low-alloyed steels frequently detached from the metallic core and spallation is observed. Besides, the chlorine model shown above the material exposed in this study suffered from the formation of alkali iron tri-sulphates, however with much lower aggressiveness as was observed under deposit D1. Potassium chloride at high temperatures undergoes decomposition to free potassium and chlorine and these elements react directly with the formed oxide scales; in lowalloyed steels with Fe 3 O 4 (chlorine cycle), whereas in the alloy with higher Cr content reaction can lead to the formation of chromates: KCl ðsþ þ Cr 2 O 3 ðsþ þ H 2 O ðgþ þ 3 4 O 2! K 2 CrO 4 þ 2HCl ðgþ ð15þ The formation of stable chromates compounds decreases the total concentration of Cr in the oxide scale leading to the formation of much poorer scales with lower corrosion resistance due to lower Cr content. 4.6 Alloys with deposit D5 Chemically deposit D5 showed similar composition to D2, D3 and D4, where high content of Al 2 O 3 and SiO 2 with 10% of Na 2 SO 4 and lack of K 2 SO 4. Deposit D5 contained CaCO 3 instead of CaO presented in D4. The corrosion degradation rate under deposit D5 was similar to deposits under D2 D4 based on kinetic and metal loss data.. Similar to other deposits high temperature such as C invoked dehydration of kaolinite producing disordered meta kaolinite Al 2 Si 2 O 7 [25]. The addition of CaCO 3 combined with Al 2 O 3.2SiO 2.2H 2 O led to reduction of active alkali flux in the deposit based on the results achieved in this work. It needs to be pointed out, that CaCO 3 (limestone) in presence of SO 2 (flue gas at high concentration) forming calcium sulphate and carbon dioxide according to the following reaction: CaCO 3 ðsþ þ SO 2 ðgþ þ 1 2 O 2ðgÞ! CaSO 4 ðsþ þ CO 2 ðgþ ð16þ Comparable to D3 deposit, in D5 deposit CaSO 4 reacts with sodium sulphate (Na 2 SO 4 ) as shown below: 2CaSO 4ðsÞ þ Na 2 SO 4ðsÞ! Na 2 Ca 2 ðso 4 Þ 3ðsÞ ð17þ The sodium calcium tri-sulphate Na 2 Ca 2 (SO 4 ) 3 based on kinetics and the metal loss data is less corrosive than K 3 Fe(SO 4 ) 3 and Na 3 Fe(SO 4 ) 3 formed under deposit D1. Furthermore, much lower mass gain of the low-alloyed steels was found because of the high concentration of kaolinite added to decrease corrosive behaviour of 10% of Na 2 SO 4 in D5 deposit [10]. 5 Conclusions The aim of this study was to investigate the corrosion behaviour of four ferritic alloys: 15Mo3, T22, T23 and T91 in simulated cofired combustion gases for 1000 h at C with different chemistry of the deposits. Dimensional metrology was used as the primary route to quantify the material damage and to produce statistically significant metal loss distributions. The crosssections of the samples were examined in ESEM/ EDX. The following conclusions can be made from this study: 1. The materials covered with deposit D1 showed the highest specific net mass gain and the highest metal loss due to the formation of alkali iron tri-sulphates. The alloys covered in deposit D1 underwent high corrosion rate due to the lack of formation of a protective oxide layer. 2. The alloys formed non-protective oxide scales consisting of Fe 2 O 3 (top layer) and Fe 3 O 4 (inner layer) when no deposit was used (deposit D0). Additional elements such as S and Cr were detected in the scales formed on T22; S, Cr and W in T23 and Cr 2 S 3 in T91. Thick scales were formed on 15Mo3, T22 and T23, much thinner scale was observed in T91 due to higher Cr content. Nevertheless, the scales showed poor adherence to the substrate. 3. Similar degradation rate was found under deposit D2 and D3 due to the similar composition of both deposits. Under D3 deposit less aggressive alkali sulphates: Na 2 Ca 2 (SO 4 ) 3 and K 2 Ca 2 (SO 4 ) 3 formed. The dimensional metal loss data indicates the corrosion damage of 15Mo3, T22 and T23 was similar under D4 and D5 deposits. 4. Corrosion degradation decreased with increasing Cr content in the alloys for all six deposits. The median metal damage of the alloys showed the following ranking (most to the least damage): 15Mo3> T22> T23 >T91. 6 References [1] International Energy Agency (IEA)-Report, CO 2 emissions from fossil fuel combustion, IEA Statistics, [2] K. Natesan, J. H. Park, Int. J. Hydrogen Energy. 2007, 32, [3] S. Hastia, A. Aroonwilasa, A. Veawaba, Energy Procedia. 2013, 37, [4] J. Skea, P. Ekins., M. Winskel, Making the transition to a secure and low-carbon energy system, Routledge, London [5] A. A. Khan, W. de Jong, P. J. Jansens, H. Spliethoff., Fuel Process Technol. 2009, 90, 21. [6] J. Henry, G. Zhou, T. Ward., Mater High Temp. 2007, 24, 249. [7] A. U. Syed, N. J. Simms, J. E. Oakey, Fuel. 2012, 101, 62. [8] T. Hussain, A. U. Syed, N. J. Simms, Fuel. 2013, 113, 787. [9] K. Natesan, A. Purohit, D. L. Rink, Proc. of 17th Annual Conference on Fossil Energy Materials. Baltimore, United States April [10] T. Dudziak, T. Hussain, N. J. Simms, A. U. Syed, J. E. Oakey, Corros. Sci. 2014, 79, 184. [11] N. J. Simms, A. T. Fry., Modelling fireside corrosion of heat exchangers in co-fired pulverised fuel power systems. In: Lecomte-Beckers J, Carton M (eds.) Materials for advanced power engineering. Julich: Forschungszentrum;Germany, WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

12 850 Dudziak, Hussain, Orlicka, Pokrywa and Simms Materials and Corrosion 2015, 66, No. 9 [12] N. J. Simms, P. J. Kilgallon, J. E. Oakey, Mater Sci. Eng Energy Syst. 2007, 2, 154. [13] C. W. Bale, A. D. Pelton, W. J. Thompson, G. Eriksson, K. Hack, P. Chartrand, S. Decterov, Jung, J. Melancon, S. Petersen, FactSage 6.2 (database: FACT), Copyright Thermfact and GTT-Technologies [14] A. D. Smigelskas, E. O. Kirkendall, Trans. AIME 1947, 171, 130. [15] J. Ellis, M. Haw, Chromium Carbides, Materials World 1997, 5 (11), 1. [16] P. Huczkowski, T. Olszewski, M. Schiek, B. Lutz, G. R. Holcomb, V. Shemet, W. Nowak, G. H. Meier, L. Singheiser, W. J. Quadakkers, Mater. Corros. 2014, 65, 2. [17] J. Stringer, I. G. Wright, Oxi. Met. 1995, 44, 265. [18] A. Syed, PhD Thesis, Cranfield University, Cranfield, United Kingdom [19] S. Kihara, T. Isozaki, A. Ohtomo, Proc. JIMIS 3 High- Temperature Corrosion, Tokyo, Japan [20] A. Karlsson, P. J. Møller, V. Johansen, Corros. Sci. 1990, 30, 153. [21] F. Mansfeld, U. Bertocci, Electrochemical Corrosion Testing, ASTM International, Baltimore [22] K. Natesan, J. Met. 1991, 43, 61. [23] W. T. Reid, External Corrosion and Deposits, American Elsevier, New York [24] M. Bellotto, A. Gualtieri, G. Artioli, S. M. Clark, Phys. Chem. Minerals. 1995, 22, 207. [25] M. Aho, J. Silvennoinen, Fuel 2004, 84, [26] P. Viklund, PhD Thesis, KTH, Stockholm, Sweden [27] U. Nieken, O. Watzenburger, Chem. Eng. Sci. 1999, 54, [28] E. Reese, H. J. Grabke, Mater. Corros. 1992, 43, 547. [29] E. Reese, H. J. Grabke, Mater. Corros. 1993, 44, 41. (Received: July 4, 2014) (Accepted: December 2, 2014) W WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim